Modeling induction phenomena in intermolecular interactions with an ab lnitio force field

One possible road toward the development of a polarizable potential energy function relies on the use of distributed polarizabilities derived from the induction energy mapped around the molecule. Whereas such polarizable models are expected to reproduce the signature induction energy with an appreciable accuracy, it is far from clear whether they will perform equally well in the context of intermolecular interactions. To address this issue, while pursuing the ultimate goal of a "plug-and-play"-like approach, polarizability models determined quantum mechanically and consisting of atomic isotropic dipole plus charge-flow polarizabilities were combined with the classical, nonpolarizable Charmm force field. Performance of the models was probed in the challenging test cases of cation-π binding and the association of a divalent calcium ion with water, where induction effects are envisioned to be considerable. Since brute force comparison of the binding energies estimated from the polarizable and the classical Charmm potential energy functions is not justified, the individual electrostatic and induction contributions of the force field were confronted to the corresponding terms of a symmetry-adapted perturbation theory (SAPT) expansion carried out with the 6-311 ++G(d,p) basis set. While the quantum-mechanical and the molecular-mechanical electrostatic and damped induction contributions agree reasonably well, overall reproduction of the binding energies is plagued by an underestimated repulsion that underlines the necessity of de novo parametrization of the classical 6-12 form of the van der Waals potential. Based on the SAPT expansion, new Lennard-Jones parameters were optimized, which, combined with the remainder of the polarizable force field, yield an improved reproduction of the target binding energies.